基于改进孤立森林算法的电力营销异常数据识别方法研究
Modified Isolation Forest-based Abnormal Data Identification in Electric Marketing
王宏宇 1李学安1
作者信息
- 1. 凌海供电有限责任公司,辽宁锦州 121001
- 折叠
摘要
针对现有识别方法识别电力营销数据时存在识别结果的AUC值过低,无法满足电力企业对异常数据识别的性能要求问题,引入改进孤立森林算法,开展对电力营销异常数据识别方法的研究.采集电力营销数据,并对采集数据进行融合;提取电力营销数据特征,并将数据分解成各个模态分量;利用改进孤立森林算法,实现数据的异常识别.通过对比实验证明:新识别方法在实际应用中得到的识别结果AUC值更接近1,具备极高的识别性能,值得广泛应用和推广.
Abstract
In view of the problem that the AUC value of the recognition result is too low when the currently prevailing i-dentification methods are used to identify power marketing data,which cannot meet the performance requirements of pow-er enterprises for abnormal data recognition,this paper introduces a modified isolation forest algorithm to study the abnor-mal data recognition method of power marketing.By collecting and fusing power marketing data,the features of power marketing data are extracted and decomposed into each modal component.Using the modified isolation forest algorithm,the data anomaly identification is realized.Through comparison experiment,it is proved that the AUC value of the new recognition method is closer to 1 in practical application,which has high identification performance and is worthy of wide application and popularization.
关键词
改进孤立森林算法/营销/数据识别/异常/电力Key words
modified isolation forest algorithm/marketing/data identification/abnormality/electrics引用本文复制引用
出版年
2024
NETL
NSTL
万方数据